Error: An input is not a numeric array (or GPU support not compiled).

[eduardo4jesus]

I am getting this error when running cnn_fft_train.m

An input is not a numeric array (or GPU support not compiled).

Error in dagnn.Conv/forward (line 11)
      outputs{1} = vl_nnconv(...

Error in dagnn.Layer/forwardAdvanced (line 85)
      outputs = obj.forward(inputs, {net.params(par).value}) ;

Error in dagnn.DagNN/eval (line 91)
  obj.layers(l).block.forwardAdvanced(obj.layers(l)) ;

Error in cnn_train_dag>processEpoch (line 253)
      net.eval(inputs, params.derOutputs, 'holdOn', s < params.numSubBatches) ;

Error in cnn_train_dag (line 105)
    [net, state] = processEpoch(net, state, params, 'train') ;

Error in cnn_fft_train (line 72)
[net, info] = trainfn(net, imdb, getBatch(opts), ...

---

Would you have any tips on how to sort this out?

[eduardo4jesus]

I tried to compile MatConvNet with CUDA, but no success on that.

>> vl_compilenn('enableGpu', true, 'cudaMethod', 'nvcc', ...
'cudaRoot', '/usr/local/cuda', ...
'enableCudnn', true, 'cudnnRoot', '/usr/local/cudnn')
vl_compilenn:   CUDA: MEX config file: '/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/config/mex_CUDA_glnxa64.xml'
/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(69): error: identifier "cudnnConvolutionFwdPreference_t" is undefined
      void setCudnnConvolutionFwdPreference(cudnnConvolutionFwdPreference_t x,
                                            ^

/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(74): error: identifier "cudnnConvolutionBwdFilterPreference_t" is undefined
      void setCudnnConvolutionBwdFilterPreference(cudnnConvolutionBwdFilterPreference_t x,
                                                  ^

/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(79): error: identifier "cudnnConvolutionBwdDataPreference_t" is undefined
      void setCudnnConvolutionBwdDataPreference(cudnnConvolutionBwdDataPreference_t x,
                                                ^

/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(120): error: identifier "cudnnConvolutionFwdPreference_t" is undefined
      cudnnConvolutionFwdPreference_t cudnnConvolutionFwdPreference ;
      ^

/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(126): error: identifier "cudnnConvolutionBwdFilterPreference_t" is undefined
      cudnnConvolutionBwdFilterPreference_t cudnnConvolutionBwdFilterPreference;
      ^

/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/datacu.hpp(132): error: identifier "cudnnConvolutionBwdDataPreference_t" is undefined
      cudnnConvolutionBwdDataPreference_t cudnnConvolutionBwdDataPreference ;
      ^

6 errors detected in the compilation of "/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/data.cu".
Error using vl_compilenn>nvcc_compile (line 540)
Command "/usr/local/cuda/bin/nvcc" -c "/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/src/bits/data.cu"
-DNDEBUG -DENABLE_GPU -DENABLE_CUDNN -I"/usr/local/cudnn/include" -DENABLE_DOUBLE -gencode=arch=compute_86,code=\"sm_86,compute_86\"
-I"/home/eduardoj/MATLAB/R2024a/extern/include" -I"/home/eduardoj/MATLAB/R2024a/toolbox/distcomp/gpu/extern/include"  -Xcompiler
-mssse3,-ftree-vect-loop-version,-ffast-math,-funroll-all-loops -gencode=arch=compute_86,code=\"sm_86,compute_86\"  -O3 -Xcompiler -fPIC -D_FORCE_INLINES -o
"/home/eduardoj/Code/junxuan-li/A-frequency-domain-neural-network-for-fast-image-super-resolution/matlab/mex/.build/bits/data.o" failed.

Error in vl_compilenn (line 485)
      nvcc_compile(opts, srcs{i}, objfile, flags.nvcc) ;